hi Julian,

you may want to have a look at the ann library:

• ann_som: Self Organizing Maps
• ann_mlp: multi layered perceptrons (feed forward)
• ann_td: time delay nn (currently a bit bugged)

you can find the source code in the CVS: http://cvs.sourceforge.net/viewcvs.py/pure-data/externals/ann/ see this page for a quick cvs howto: http://sourceforge.net/cvs/?group_id=55736

ann_mlp and ann_td use the fann library, so you'll need it on your system to compile ann. see: http://leenissen.dk/fann/ there you can find info on training algorithms etc..

ann works on win, osx and linux.

...in the spare time I'm trying to extend ann, I'd like to include RNN or even something to let the user define the topology without limitations...

if you have questions let me know.

ciao and welcome in the pd community! davide.

----- Original Message ----- From: Julian Villegas To: pd-list@iem.at Sent: Wednesday, March 08, 2006 6:17 PM Subject: [PD] neural networks & PD

I'm new in PD and I will like to know if someone have worked before with neural networks in pd, and how fast they can be. I found something in La kitchen, and in some papers, but any 'light' on how to do it and how fast they run, or whatever I should be aware before to start will be highly appreciate it ! :) Thanks in advance

Julian Villegas

University of Aizu Synthetic Worlds Laboratory Aizu - Wakamatsu Fukushima Japan 965 - 8580

Me pregunto de un modo pensativo Que significa ser colombiano? No se le respondi. Es un acto de fe JLB.

Yahoo! Mail Bring photos to life! New PhotoMail makes sharing a breeze.

PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list